Utilizing the database of seismic landslides and strong motion records from the Loess Plateau, 81 earthquakes with magnitudes ranging from 3.0 to 7.1 were undertaken. Furthermore, 2802 strong ground motion records were chosen as the focal points of our investigation. Through analysis of the variation pattern of effective peak acceleration (EPA) values, an EPA prediction model for the Loess Plateau was developed using both surface wave magnitude (Ms) and Das magnitude (Mwg). Furthermore, the study examines the discrepancies between this model and two other attenuation models. The results show that (1) the \(\overline{\beta }\) max (maximum average amplification factor) values employed for the Loess Plateau differ from those specified in other established standards and span from 2.15 to 2.40 in the horizontal and vertical directions. (2) A negative correlation between EPA values and epicentral distance. The magnitudes and directions influence the rates of change of the fitting curves. (3) A positive correlation between the EPA values and earthquake magnitudes. The change rates of the fitting curves tend to increase first and then decrease, with the horizontal directions always being higher than the vertical directions. (4) The prediction equation based on Mwg exhibits residual standard deviations in the range of 0.493–0.536, which optimizes the fitting results and simplifies the fitting procedure to some extent. (5) The prediction model for the Loess Plateau differs from the reference models, with its fitting EPA values and attenuation rates being relatively lower.
